Abstract – There has been amazing progress in computer vision, machine learning, and AI over the past few years. State of the art visual recognition methods can now reliably perform object recognition and detection. Despite this amazing progress, there is still a large gap between the level of visual understanding in intelligent agents in real world (animals, infants, etc.) and that of our best algorithms. In this talk, I argue for visual reasoning as the next crucial step in computer vision and AI. Reasoning and planning in a visual world requires two main components: visual knowledge, and reasoning machinery. I start this talk by showing examples of visual knowledge extraction methods, followed by some of our steps toward visual reasoning.
Bio – Ali Farhadi is an Assistant Professor in the Department of Computer Science & Engineering at the University of Washington. He also leads the project Plato at Allen Institute for Artificial Intelligence in Seattle. Prior to this, he was a postdoctoral fellow at the Robotics Institute at Carnegie Mellon University. He received his PhD from University of Illinois at Urbana-Champaign under the supervision of David Forsyth. His research has been mainly focused on computer vision and machine learning. More specifically, he is interested in semantic scene understanding, visual knowledge extraction, object recognition, transfer learning, and attribute based representations of objects. Ali has been awarded the Allen Distinguished Investigator Award, the inaugural Google fellowship in computer vision and image interpretation, the university of Illinois CS/AI 2009 award, C.W. Gear 2010 Outstanding Research Award, and CVPR2011 and AAAI 2016 Best Student Paper Award.
Sponsored by the Laboratory for Analytic Sciences, The Department of Electrical and Computer Engineering, and the Army Research Office.
LAS aims to bring together a multi-disciplinary group of academic, industry, and government researchers, analysts and managers together to re-engineer the intelligence analysis process to promote predictive analysis. LAS will do this by conducting both classified and unclassified research in a variety of areas of research. The research done in this area will serve as the foundation for mission effects and integrated back into the enterprise.
If you would like to sign up for our email distribution list, please fill out this form: